wzry-vits-api / README_zh.md
Artrajz's picture
Upload 103 files
c5ed230
|
raw
history blame
24.9 kB

vits-simple-api

Simply call the vits api


English|中文文档

Feature

  • VITS语音合成
  • VITS语音转换
  • HuBert-soft VITS模型
  • W2V2 VITS / emotional-vits维度情感模型
  • 加载多模型
  • 自动识别语言并处理,根据模型的cleaner设置语言类型识别的范围,支持自定义语言类型范围
  • 自定义默认参数
  • 长文本批处理
  • GPU加速推理
  • SSML语音合成标记语言(完善中...)
Update Logs

2023.5.24

添加dimensional_emotion api,从文件夹加载多个npy文件,Docker添加了Linux/ARM64和Linux/ARM64/v8平台

2023.5.15

增加english_cleaner,需要额外安装espeak才能使用

2023.5.12

增加ssml支持,但仍需完善。重构部分功能,hubert_vits中的speaker_id改为id

2023.5.2

增加w2v2-vits/emotional-vits模型支持,修改了speakers映射表并添加了对应模型支持的语言

2023.4.23

增加api key鉴权,默认禁用,需要在config.py中启用

2023.4.17

修改单语言的cleaner需要标注才会clean,增加GPU加速推理,但需要手动安装gpu推理环境

2023.4.12

项目由MoeGoe-Simple-API更名为vits-simple-api,支持长文本批处理,增加长文本分段阈值max

2023.4.7

增加配置文件可自定义默认参数,本次更新需要手动更新config.py,具体使用方法见config.py

2023.4.6

加入自动识别语种选项auto,lang参数默认修改为auto,自动识别仍有一定缺陷,请自行选择

统一POST请求类型为multipart/form-data

demo

  • https://api.artrajz.cn/py/voice/vits?text=你好,こんにちは&id=142
  • 激动:https://api.artrajz.cn/py/voice/w2v2-vits?text=こんにちは&id=3&emotion=111
  • 小声:https://api.artrajz.cn/py/voice/w2v2-vits?text=こんにちは&id=3&emotion=2077

https://user-images.githubusercontent.com/73542220/237995061-c1f25b4e-dd86-438a-9363-4bb1fe65b425.mov

demo服务器配置比较低所以不稳定

部署

Docker部署

镜像拉取脚本

bash -c "$(wget -O- https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/vits-simple-api-installer-latest.sh)"
  • 目前docker镜像支持的平台linux/amd64,linux/arm64
  • 在拉取完成后,需要导入VITS模型才能使用,请根据以下步骤导入模型。

下载VITS模型

将模型放入/usr/local/vits-simple-api/Model

Folder structure

│  hubert-soft-0d54a1f4.pt
│  model.onnx
│  model.yaml
├─g
│      config.json
│      G_953000.pth
│
├─louise
│      360_epochs.pth
│      config.json
│
├─Nene_Nanami_Rong_Tang
│      1374_epochs.pth
│      config.json
│
├─Zero_no_tsukaima
│       1158_epochs.pth
│       config.json
│
└─npy
       25ecb3f6-f968-11ed-b094-e0d4e84af078.npy
       all_emotions.npy

修改模型路径

Modify in /usr/local/vits-simple-api/config.py

config.py

# 在此填写模型路径
MODEL_LIST = [
    # VITS
    [ABS_PATH + "/Model/Nene_Nanami_Rong_Tang/1374_epochs.pth", ABS_PATH + "/Model/Nene_Nanami_Rong_Tang/config.json"],
    [ABS_PATH + "/Model/Zero_no_tsukaima/1158_epochs.pth", ABS_PATH + "/Model/Zero_no_tsukaima/config.json"],
    [ABS_PATH + "/Model/g/G_953000.pth", ABS_PATH + "/Model/g/config.json"],
    # HuBert-VITS (Need to configure HUBERT_SOFT_MODEL)
    [ABS_PATH + "/Model/louise/360_epochs.pth", ABS_PATH + "/Model/louise/config.json"],
    # W2V2-VITS (Need to configure DIMENSIONAL_EMOTION_NPY)
    [ABS_PATH + "/Model/w2v2-vits/1026_epochs.pth", ABS_PATH + "/Model/w2v2-vits/config.json"],
]
# hubert-vits: hubert soft 编码器
HUBERT_SOFT_MODEL = ABS_PATH + "/Model/hubert-soft-0d54a1f4.pt"
# w2v2-vits: Dimensional emotion npy file
# 加载单独的npy: ABS_PATH+"/all_emotions.npy
# 加载多个npy: [ABS_PATH + "/emotions1.npy", ABS_PATH + "/emotions2.npy"]
# 从文件夹里加载npy: ABS_PATH + "/Model/npy"
DIMENSIONAL_EMOTION_NPY = ABS_PATH + "/Model/npy"
# w2v2-vits: 需要在同一路径下有model.onnx和model.yaml
DIMENSIONAL_EMOTION_MODEL = ABS_PATH + "/Model/model.yaml"

启动

docker compose up -d

或者重新执行拉取脚本

镜像更新

重新执行docker镜像拉取脚本即可

虚拟环境部署

Clone

git clone https://github.com/Artrajz/vits-simple-api.git

下载python依赖

推荐使用python的虚拟环境,python版本 >= 3.9

pip install -r requirements.txt

windows下可能安装不了fasttext,可以用以下命令安装,附wheels下载地址

#python3.10 win_amd64
pip install https://github.com/Artrajz/archived/raw/main/fasttext/fasttext-0.9.2-cp310-cp310-win_amd64.whl
#python3.9 win_amd64
pip install https://github.com/Artrajz/archived/raw/main/fasttext/fasttext-0.9.2-cp39-cp39-win_amd64.whl

下载VITS模型

将模型放入 /path/to/vits-simple-api/Model

文件夹结构

├─g
│      config.json
│      G_953000.pth
│
├─louise
│      360_epochs.pth
│      config.json
│      hubert-soft-0d54a1f4.pt
│
├─Nene_Nanami_Rong_Tang
│      1374_epochs.pth
│      config.json
│
└─Zero_no_tsukaima
        1158_epochs.pth
        config.json

修改模型路径

/path/to/vits-simple-api/config.py 修改

config.py

# 在此填写模型路径
MODEL_LIST = [
    # VITS
    [ABS_PATH + "/Model/Nene_Nanami_Rong_Tang/1374_epochs.pth", ABS_PATH + "/Model/Nene_Nanami_Rong_Tang/config.json"],
    [ABS_PATH + "/Model/Zero_no_tsukaima/1158_epochs.pth", ABS_PATH + "/Model/Zero_no_tsukaima/config.json"],
    [ABS_PATH + "/Model/g/G_953000.pth", ABS_PATH + "/Model/g/config.json"],
    # HuBert-VITS (Need to configure HUBERT_SOFT_MODEL)
    [ABS_PATH + "/Model/louise/360_epochs.pth", ABS_PATH + "/Model/louise/config.json"],
    # W2V2-VITS (Need to configure DIMENSIONAL_EMOTION_NPY)
    [ABS_PATH + "/Model/w2v2-vits/1026_epochs.pth", ABS_PATH + "/Model/w2v2-vits/config.json"],
]
# hubert-vits: hubert soft 编码器
HUBERT_SOFT_MODEL = ABS_PATH + "/Model/hubert-soft-0d54a1f4.pt"
# w2v2-vits: Dimensional emotion npy file
# 加载单独的npy: ABS_PATH+"/all_emotions.npy
# 加载多个npy: [ABS_PATH + "/emotions1.npy", ABS_PATH + "/emotions2.npy"]
# 从文件夹里加载npy: ABS_PATH + "/Model/npy"
DIMENSIONAL_EMOTION_NPY = ABS_PATH + "/Model/npy"
# w2v2-vits: 需要在同一路径下有model.onnx和model.yaml
DIMENSIONAL_EMOTION_MODEL = ABS_PATH + "/Model/model.yaml"

启动

python app.py

GPU 加速

windows

安装CUDA

查看显卡最高支持CUDA的版本

nvidia-smi

以CUDA11.7为例,官网

安装GPU版pytorch

CUDA11.7对应的pytorch是用这个命令安装

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

对应版本的命令可以在官网找到

Linux

安装过程类似,但我没有相应的环境所以没办法测试

Openjtalk安装问题

如果你是arm64架构的平台,由于pypi官网上没有arm64对应的whl,可能安装会出现一些问题,你可以使用我构建的whl来安装

pip install openjtalk==0.3.0.dev2 --index-url https://pypi.artrajz.cn/simple

或者是自己手动构建一个whl,可以根据教程来构建

API

GET

speakers list

voice vits

check

POST

  • python
import re
import requests
import os
import random
import string
from requests_toolbelt.multipart.encoder import MultipartEncoder

abs_path = os.path.dirname(__file__)
base = "http://127.0.0.1:23456"


# 映射表
def voice_speakers():
    url = f"{base}/voice/speakers"

    res = requests.post(url=url)
    json = res.json()
    for i in json:
        print(i)
        for j in json[i]:
            print(j)
    return json


# 语音合成 voice vits
def voice_vits(text, id=0, format="wav", lang="auto", length=1, noise=0.667, noisew=0.8, max=50):
    fields = {
        "text": text,
        "id": str(id),
        "format": format,
        "lang": lang,
        "length": str(length),
        "noise": str(noise),
        "noisew": str(noisew),
        "max": str(max)
    }
    boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))

    m = MultipartEncoder(fields=fields, boundary=boundary)
    headers = {"Content-Type": m.content_type}
    url = f"{base}/voice"

    res = requests.post(url=url, data=m, headers=headers)
    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path


# 语音转换 hubert-vits
def voice_hubert_vits(upload_path, id, format="wav", length=1, noise=0.667, noisew=0.8):
    upload_name = os.path.basename(upload_path)
    upload_type = f'audio/{upload_name.split(".")[1]}'  # wav,ogg

    with open(upload_path, 'rb') as upload_file:
        fields = {
            "upload": (upload_name, upload_file, upload_type),
            "id": str(id),
            "format": format,
            "length": str(length),
            "noise": str(noise),
            "noisew": str(noisew),
        }
        boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))

        m = MultipartEncoder(fields=fields, boundary=boundary)
        headers = {"Content-Type": m.content_type}
        url = f"{base}/voice/hubert-vits"

        res = requests.post(url=url, data=m, headers=headers)
    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path


# 维度情感模型 w2v2-vits
def voice_w2v2_vits(text, id=0, format="wav", lang="auto", length=1, noise=0.667, noisew=0.8, max=50, emotion=0):
    fields = {
        "text": text,
        "id": str(id),
        "format": format,
        "lang": lang,
        "length": str(length),
        "noise": str(noise),
        "noisew": str(noisew),
        "max": str(max),
        "emotion": str(emotion)
    }
    boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))

    m = MultipartEncoder(fields=fields, boundary=boundary)
    headers = {"Content-Type": m.content_type}
    url = f"{base}/voice/w2v2-vits"

    res = requests.post(url=url, data=m, headers=headers)
    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path


# 语音转换 同VITS模型内角色之间的音色转换
def voice_conversion(upload_path, original_id, target_id):
    upload_name = os.path.basename(upload_path)
    upload_type = f'audio/{upload_name.split(".")[1]}'  # wav,ogg

    with open(upload_path, 'rb') as upload_file:
        fields = {
            "upload": (upload_name, upload_file, upload_type),
            "original_id": str(original_id),
            "target_id": str(target_id),
        }
        boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))
        m = MultipartEncoder(fields=fields, boundary=boundary)

        headers = {"Content-Type": m.content_type}
        url = f"{base}/voice/conversion"

        res = requests.post(url=url, data=m, headers=headers)

    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path


def voice_ssml(ssml):
    fields = {
        "ssml": ssml,
    }
    boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))

    m = MultipartEncoder(fields=fields, boundary=boundary)
    headers = {"Content-Type": m.content_type}
    url = f"{base}/voice/ssml"

    res = requests.post(url=url, data=m, headers=headers)
    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path

def voice_dimensional_emotion(upload_path):
    upload_name = os.path.basename(upload_path)
    upload_type = f'audio/{upload_name.split(".")[1]}'  # wav,ogg

    with open(upload_path, 'rb') as upload_file:
        fields = {
            "upload": (upload_name, upload_file, upload_type),
        }
        boundary = '----VoiceConversionFormBoundary' + ''.join(random.sample(string.ascii_letters + string.digits, 16))

        m = MultipartEncoder(fields=fields, boundary=boundary)
        headers = {"Content-Type": m.content_type}
        url = f"{base}/voice/dimension-emotion"

        res = requests.post(url=url, data=m, headers=headers)
    fname = re.findall("filename=(.+)", res.headers["Content-Disposition"])[0]
    path = f"{abs_path}/{fname}"

    with open(path, "wb") as f:
        f.write(res.content)
    print(path)
    return path

API KEY

在config.py中设置API_KEY_ENABLED = True以启用,api key填写:API_KEY = "api-key"

启用后,GET请求中使用需要增加参数api_key,POST请求中使用需要在header中添加参数X-API-KEY

Parameter

VITS语音合成

Name Parameter Is must Default Type Instruction
合成文本 text true str
角色id id false 0 int
音频格式 format false wav str wav,ogg,silk
文本语言 lang false auto str auto为自动识别语言模式,也是默认模式。lang=mix时,文本应该用[ZH] 或 [JA] 包裹。方言无法自动识别。
语音长度/语速 length false 1.0 float 调节语音长度,相当于调节语速,该数值越大语速越慢
噪声 noise false 0.667 float
噪声偏差 noisew false 0.8 float
分段阈值 max false 50 int 按标点符号分段,加起来大于max时为一段文本。max<=0表示不分段。

VITS 语音转换

Name Parameter Is must Default Type Instruction
上传音频 upload true file wav or ogg
源角色id original_id true int 上传文件所使用的角色id
目标角色id target_id true int 要转换的目标角色id

HuBert-VITS 语音转换

Name Parameter Is must Default Type Instruction
上传音频 upload true file
目标角色id id true int
音频格式 format true str wav,ogg,silk
语音长度/语速 length true float 调节语音长度,相当于调节语速,该数值越大语速越慢
噪声 noise true float
噪声偏差 noisew true float

Dimensional emotion

Name Parameter Is must Default Type Instruction
上传音频 upload true file 返回存储维度情感向量的npy文件

W2V2-VITS

Name Parameter Is must Default Type Instruction
合成文本 text true str
角色id id false 0 int
音频格式 format false wav str wav,ogg,silk
文本语言 lang false auto str auto为自动识别语言模式,也是默认模式。lang=mix时,文本应该用[ZH] 或 [JA] 包裹。方言无法自动识别。
语音长度/语速 length false 1.0 float 调节语音长度,相当于调节语速,该数值越大语速越慢
噪声 noise false 0.667 float
噪声偏差 noisew false 0.8 float
分段阈值 max false 50 int 按标点符号分段,加起来大于max时为一段文本。max<=0表示不分段。
维度情感 emotion false 0 int 范围取决于npy情感参考文件,如innnky的all_emotions.npy模型范围是0-5457

SSML语音合成标记语言

目前支持的元素与属性

speak元素

Attribute Description Is must
id 默认值从config.py中读取 false
lang 默认值从config.py中读取 false
length 默认值从config.py中读取 false
noise 默认值从config.py中读取 false
noisew 默认值从config.py中读取 false
max 按标点符号分段,加起来大于max时为一段文本。max<=0表示不分段,这里默认为0。 false
model 默认为vits,可选w2v2-vitsemotion-vits false
emotion 只有用w2v2-vitsemotion-vitsemotion才生效,范围取决于npy情感参考文件 false

voice元素

优先级大于speak

Attribute Description Is must
id 默认值从config.py中读取 false
lang 默认值从config.py中读取 false
length 默认值从config.py中读取 false
noise 默认值从config.py中读取 false
noisew 默认值从config.py中读取 false
max 按标点符号分段,加起来大于max时为一段文本。max<=0表示不分段,这里默认为0。 false
model 默认为vits,可选w2v2-vitsemotion-vits false
emotion 只有用w2v2-vitsemotion-vitsemotion才会生效 false

break元素

Attribute Description Is must
strength x-weak,weak,medium(默认值),strong,x-strong false
time 暂停的绝对持续时间,以秒为单位(例如 2s)或以毫秒为单位(例如 500ms)。 有效值的范围为 0 到 5000 毫秒。 如果设置的值大于支持的最大值,则服务将使用 5000ms。 如果设置了 time 属性,则会忽略 strength 属性。 false
Strength Relative Duration
x-weak 250 毫秒
weak 500 毫秒
Medium 750 毫秒
Strong 1000 毫秒
x-strong 1250 毫秒

示例

<speak lang="zh" format="mp3" length="1.2">
    <voice id="92" >这几天心里颇不宁静。</voice>
    <voice id="125">今晚在院子里坐着乘凉,忽然想起日日走过的荷塘,在这满月的光里,总该另有一番样子吧。</voice>
    <voice id="142">月亮渐渐地升高了,墙外马路上孩子们的欢笑,已经听不见了;</voice>
    <voice id="98">妻在屋里拍着闰儿,迷迷糊糊地哼着眠歌。</voice>
    <voice id="120">我悄悄地披了大衫,带上门出去。</voice><break time="2s"/>
    <voice id="121">沿着荷塘,是一条曲折的小煤屑路。</voice>
    <voice id="122">这是一条幽僻的路;白天也少人走,夜晚更加寂寞。</voice>
    <voice id="123">荷塘四面,长着许多树,蓊蓊郁郁的。</voice>
    <voice id="124">路的一旁,是些杨柳,和一些不知道名字的树。</voice>
    <voice id="125">没有月光的晚上,这路上阴森森的,有些怕人。</voice>
    <voice id="126">今晚却很好,虽然月光也还是淡淡的。</voice><break time="2s"/>
    <voice id="127">路上只我一个人,背着手踱着。</voice>
    <voice id="128">这一片天地好像是我的;我也像超出了平常的自己,到了另一个世界里。</voice>
    <voice id="129">我爱热闹,也爱冷静;<break strength="x-weak"/>爱群居,也爱独处。</voice>
    <voice id="130">像今晚上,一个人在这苍茫的月下,什么都可以想,什么都可以不想,便觉是个自由的人。</voice>
    <voice id="131">白天里一定要做的事,一定要说的话,现在都可不理。</voice>
    <voice id="132">这是独处的妙处,我且受用这无边的荷香月色好了。</voice>
</speak>

交流平台

现在只有 Q群

鸣谢